The methodology is in the post and everything is open source — anyone can reproduce it. This dataset isn't public because at least 38 people don't know their key is exposed. Future analyses will release everything. Happy to share privately for legitimate research.